In 2022, the Federal Housing Finance Agency approved two new credit scores—FICO 10T and VantageScore 4.0—to replace the Classic FICO credit score the government-sponsored enterprises (GSEs) required for mortgage origination. This change intended to modernize credit scoring and to introduce competition, which could make purchasing a home affordable for more people.
With VantageScore 4.0 now in the pilot phase (and the use of FICO 10T expected to follow), lenders in the pilot can deliver either VantageScore or Classic FICO for included loans. When a borrower is scored by both FICO and VantageScore, the lender can present the more favorable score. We call this option “lender choice shopping.”
To understand how lender choice shopping affects the market’s assumed probability of default, we compare the Classic FICO and VantageScore 4.0 scores on GSE loans that originated between 2013 and 2023 for which both scores are available. We use the tri-merge median FICO score for each borrower (taking the lower score among coborrowers), and we place the VantageScore on the FICO scale by subtracting 20 points, as some market participants do. We anticipate that when the credit score modernization is complete, the GSEs will use the probability of default for each score to map across score scales.
We find that shopping a credit score to report the higher option would shift the credit score distribution upward, which would affect the pricing of the GSEs’ credit risk transfer deals and significantly reduce the GSEs’ revenue from loan-level pricing. As a result, borrowers would receive more affordable loan pricing in the short term while the GSEs would face greater uncertainty around borrower credit risk.
How lender choice shopping shifts the borrower credit score distribution
With lender choice shopping, we see a large positive shift in the credit score distribution among GSE-backed loans. The highest band of scores (800 and above) jumps from 13.8 percent of borrowers with the Classic FICO score to 24.0 percent when lenders can choose the higher of their Classic FICO score and VantageScore. Further, the bottom band (below 640) is cut roughly in half, from 1.5 to 0.8 percent. In all, 44 percent of borrowers improve their reported score through lender choice shopping, with the average score rising by about 11 points.
Because the scores are shifted upward under lender choice shopping, each band has a higher probability of default than either the Classic FICO score or VantageScore 4.0 would on its own, but the overall probability of default is unchanged because the GSEs do not approve loans based on credit scores.
The market, however, does not know to what extent lenders are shopping multiple scores. Initially, market participants will assume the worst. In our framework, they will accept the shopped score but use the Classic FICO distribution to forecast default probability. This assumption increases the market’s perception of the probability of default from 3.31 percent to 3.89 percent, an 18 percent increase, even though the borrower base is the same and the actual probability of default did not change, which will likely affect the GSEs’ pricing on credit risk transfer deals.
The effect of lender choice shopping on guarantee pricing
The GSEs price much of their credit risk through loan-level pricing adjustments (LLPAs), which are a one-time charge assessed to the lender at loan delivery and set by a loan’s representative credit score and loan-to-value ratio. The lender passes these costs to the borrower, either as points paid at closing or by building them into a higher note rate.
Because a higher credit score moves the loan into a lower-price cell of the grid, indicating less credit risk, lender shopping can result in borrowers receiving more favorable up-front pricing or a lower lending rate.
Consider a borrower with a Classic FICO score below 640 taking a $300,000 loan. Under the current risk pricing grid, the loan carries an LLPA of 2.49 percent of the balance, or $7,470. Presenting a higher VantageScore that moves the loan up to the 640–679 band lowers the LLPA to 2.18 percent, a reduction of roughly $920 on the same loan with the same underlying risk. The borrower pays less even as the guarantee carries the same expected loss.
Applying the current risk grid to each loan originated from 2013 to 2023 and weighting by the loan balance, the average LLPA falls roughly 9 basis points, from 0.78 percent of balance under Classic FICO to 0.69 percent under the shopped score. The reduction is largest in the lower score bands and near zero for the 800-and-above band, as borrowers in that band have no room to move up.
In 2025, a relatively low year for single-family mortgage originations, the GSEs originated approximately $750 billion in mortgages. A 9 basis-point reduction in revenue would equal a loss of $675 million.
What it means
The transition to two accepted scores—VantageScores and Classic FICO scores—is widely expected to have a small effect on qualification for a home loan because the GSEs do not rely on credit scores for qualification. Instead, lender choice shopping is most likely to affect the pricing in the credit risk transfer market, as market participants will not know the extent of the shopping and the amount of LLPA collected up front.
One caveat to this analysis is that we can see only loans that were originated. This work cannot speak for borrowers who might newly qualify or newly default under a framework that lets them shop.
Implementation plans call for the GSEs to disclose which score is used. For transparency, the Federal Housing Finance Agency and the GSEs should also consider requiring the disclosure of the Classic FICO equivalent to give the market information on the extent of the shopping.
In any case, it is improbable that federal agencies can prohibit highly engaged market participants from score shopping. They could theoretically require a lender to choose one score or the other and use that score for all loans, unless the borrower does not have it. That requirement would be difficult to implement. Lenders that do wholesale and correspondent business will likely find their broker and correspondent partners using both score types and would lose business if they were forced to pick one. Further, this requirement would not fix the issue, as credit score shopping could still occur at an earlier stage, with the real estate broker playing a role.
Ultimately, as the GSEs collect the results of this pilot program, they will have real data to assess how much lender choice shopping will affect LLPA collections and credit risk transfer pricing. As the GSEs learn more about these effects, it will be crucial for accurate risk assessment to communicate that information to market participants.
Let’s help communities build more secure, hopeful futures.
Today’s complex challenges demand smarter solutions. Urban brings decades of expertise to understanding the forces shaping people’s lives and the systems that support them. With rigorous analysis and hands-on guidance, we help leaders across the country design, test, and scale solutions that build pathways for greater opportunity.
Your support makes this possible.
